xref: /freebsd/sys/fs/ext2fs/ext2_bmap.c (revision 4f52dfbb8d6c4d446500c5b097e3806ec219fbd4)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)ufs_bmap.c	8.7 (Berkeley) 3/21/95
37  * $FreeBSD$
38  */
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/bio.h>
43 #include <sys/buf.h>
44 #include <sys/proc.h>
45 #include <sys/vnode.h>
46 #include <sys/mount.h>
47 #include <sys/racct.h>
48 #include <sys/resourcevar.h>
49 #include <sys/stat.h>
50 
51 #include <fs/ext2fs/inode.h>
52 #include <fs/ext2fs/fs.h>
53 #include <fs/ext2fs/ext2fs.h>
54 #include <fs/ext2fs/ext2_dinode.h>
55 #include <fs/ext2fs/ext2_extern.h>
56 #include <fs/ext2fs/ext2_mount.h>
57 
58 /*
59  * Bmap converts the logical block number of a file to its physical block
60  * number on the disk. The conversion is done by using the logical block
61  * number to index into the array of block pointers described by the dinode.
62  */
63 int
64 ext2_bmap(struct vop_bmap_args *ap)
65 {
66 	daddr_t blkno;
67 	int error;
68 
69 	/*
70 	 * Check for underlying vnode requests and ensure that logical
71 	 * to physical mapping is requested.
72 	 */
73 	if (ap->a_bop != NULL)
74 		*ap->a_bop = &VTOI(ap->a_vp)->i_devvp->v_bufobj;
75 	if (ap->a_bnp == NULL)
76 		return (0);
77 
78 	if (VTOI(ap->a_vp)->i_flag & IN_E4EXTENTS)
79 		error = ext4_bmapext(ap->a_vp, ap->a_bn, &blkno,
80 		    ap->a_runp, ap->a_runb);
81 	else
82 		error = ext2_bmaparray(ap->a_vp, ap->a_bn, &blkno,
83 		    ap->a_runp, ap->a_runb);
84 	*ap->a_bnp = blkno;
85 	return (error);
86 }
87 
88 /*
89  * Convert the logical block number of a file to its physical block number
90  * on the disk within ext4 extents.
91  */
92 int
93 ext4_bmapext(struct vnode *vp, int32_t bn, int64_t *bnp, int *runp, int *runb)
94 {
95 	struct inode *ip;
96 	struct m_ext2fs *fs;
97 	struct ext4_extent_header *ehp;
98 	struct ext4_extent *ep;
99 	struct ext4_extent_path *path = NULL;
100 	daddr_t lbn;
101 	int error, depth;
102 
103 	ip = VTOI(vp);
104 	fs = ip->i_e2fs;
105 	lbn = bn;
106 	ehp = (struct ext4_extent_header *)ip->i_data;
107 	depth = ehp->eh_depth;
108 
109 	*bnp = -1;
110 	if (runp != NULL)
111 		*runp = 0;
112 	if (runb != NULL)
113 		*runb = 0;
114 
115 	error = ext4_ext_find_extent(ip, lbn, &path);
116 	if (error)
117 		return (error);
118 
119 	ep = path[depth].ep_ext;
120 	if(ep) {
121 		if (lbn < ep->e_blk) {
122 			if (runp != NULL)
123 				*runp = ep->e_blk - lbn - 1;
124 		} else if (ep->e_blk <= lbn && lbn < ep->e_blk + ep->e_len) {
125 			*bnp = fsbtodb(fs, lbn - ep->e_blk +
126 			    (ep->e_start_lo | (daddr_t)ep->e_start_hi << 32));
127 			if (runp != NULL)
128 				*runp = ep->e_len - (lbn - ep->e_blk) - 1;
129 			if (runb != NULL)
130 				*runb = lbn - ep->e_blk;
131 		} else {
132 			if (runb != NULL)
133 				*runb = ep->e_blk + lbn - ep->e_len;
134 		}
135 	}
136 
137 	ext4_ext_path_free(path);
138 
139 	return (error);
140 }
141 
142 /*
143  * Indirect blocks are now on the vnode for the file.  They are given negative
144  * logical block numbers.  Indirect blocks are addressed by the negative
145  * address of the first data block to which they point.  Double indirect blocks
146  * are addressed by one less than the address of the first indirect block to
147  * which they point.  Triple indirect blocks are addressed by one less than
148  * the address of the first double indirect block to which they point.
149  *
150  * ext2_bmaparray does the bmap conversion, and if requested returns the
151  * array of logical blocks which must be traversed to get to a block.
152  * Each entry contains the offset into that block that gets you to the
153  * next block and the disk address of the block (if it is assigned).
154  */
155 
156 int
157 ext2_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, int *runp, int *runb)
158 {
159 	struct inode *ip;
160 	struct buf *bp;
161 	struct ext2mount *ump;
162 	struct mount *mp;
163 	struct indir a[EXT2_NIADDR + 1], *ap;
164 	daddr_t daddr;
165 	e2fs_lbn_t metalbn;
166 	int error, num, maxrun = 0, bsize;
167 	int *nump;
168 
169 	ap = NULL;
170 	ip = VTOI(vp);
171 	mp = vp->v_mount;
172 	ump = VFSTOEXT2(mp);
173 
174 	bsize = EXT2_BLOCK_SIZE(ump->um_e2fs);
175 
176 	if (runp) {
177 		maxrun = mp->mnt_iosize_max / bsize - 1;
178 		*runp = 0;
179 	}
180 	if (runb)
181 		*runb = 0;
182 
183 
184 	ap = a;
185 	nump = &num;
186 	error = ext2_getlbns(vp, bn, ap, nump);
187 	if (error)
188 		return (error);
189 
190 	num = *nump;
191 	if (num == 0) {
192 		*bnp = blkptrtodb(ump, ip->i_db[bn]);
193 		if (*bnp == 0) {
194 			*bnp = -1;
195 		} else if (runp) {
196 			daddr_t bnb = bn;
197 
198 			for (++bn; bn < EXT2_NDADDR && *runp < maxrun &&
199 			    is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
200 			    ++bn, ++*runp);
201 			bn = bnb;
202 			if (runb && (bn > 0)) {
203 				for (--bn; (bn >= 0) && (*runb < maxrun) &&
204 					is_sequential(ump, ip->i_db[bn],
205 						ip->i_db[bn + 1]);
206 						--bn, ++*runb);
207 			}
208 		}
209 		return (0);
210 	}
211 
212 	/* Get disk address out of indirect block array */
213 	daddr = ip->i_ib[ap->in_off];
214 
215 	for (bp = NULL, ++ap; --num; ++ap) {
216 		/*
217 		 * Exit the loop if there is no disk address assigned yet and
218 		 * the indirect block isn't in the cache, or if we were
219 		 * looking for an indirect block and we've found it.
220 		 */
221 
222 		metalbn = ap->in_lbn;
223 		if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn)
224 			break;
225 		/*
226 		 * If we get here, we've either got the block in the cache
227 		 * or we have a disk address for it, go fetch it.
228 		 */
229 		if (bp)
230 			bqrelse(bp);
231 
232 		bp = getblk(vp, metalbn, bsize, 0, 0, 0);
233 		if ((bp->b_flags & B_CACHE) == 0) {
234 #ifdef INVARIANTS
235 			if (!daddr)
236 				panic("ext2_bmaparray: indirect block not in cache");
237 #endif
238 			bp->b_blkno = blkptrtodb(ump, daddr);
239 			bp->b_iocmd = BIO_READ;
240 			bp->b_flags &= ~B_INVAL;
241 			bp->b_ioflags &= ~BIO_ERROR;
242 			vfs_busy_pages(bp, 0);
243 			bp->b_iooffset = dbtob(bp->b_blkno);
244 			bstrategy(bp);
245 #ifdef RACCT
246 			if (racct_enable) {
247 				PROC_LOCK(curproc);
248 				racct_add_buf(curproc, bp, 0);
249 				PROC_UNLOCK(curproc);
250 			}
251 #endif
252 			curthread->td_ru.ru_inblock++;
253 			error = bufwait(bp);
254 			if (error) {
255 				brelse(bp);
256 				return (error);
257 			}
258 		}
259 
260 		daddr = ((e2fs_daddr_t *)bp->b_data)[ap->in_off];
261 		if (num == 1 && daddr && runp) {
262 			for (bn = ap->in_off + 1;
263 			    bn < MNINDIR(ump) && *runp < maxrun &&
264 			    is_sequential(ump,
265 			    ((e2fs_daddr_t *)bp->b_data)[bn - 1],
266 			    ((e2fs_daddr_t *)bp->b_data)[bn]);
267 			    ++bn, ++*runp);
268 			bn = ap->in_off;
269 			if (runb && bn) {
270 				for (--bn; bn >= 0 && *runb < maxrun &&
271 					is_sequential(ump,
272 					((e2fs_daddr_t *)bp->b_data)[bn],
273 					((e2fs_daddr_t *)bp->b_data)[bn + 1]);
274 					--bn, ++*runb);
275 			}
276 		}
277 	}
278 	if (bp)
279 		bqrelse(bp);
280 
281 	/*
282 	 * Since this is FFS independent code, we are out of scope for the
283 	 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they
284 	 * will fall in the range 1..um_seqinc, so we use that test and
285 	 * return a request for a zeroed out buffer if attempts are made
286 	 * to read a BLK_NOCOPY or BLK_SNAP block.
287 	 */
288 	if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc) {
289 		*bnp = -1;
290 		return (0);
291 	}
292 	*bnp = blkptrtodb(ump, daddr);
293 	if (*bnp == 0) {
294 		*bnp = -1;
295 	}
296 	return (0);
297 }
298 
299 /*
300  * Create an array of logical block number/offset pairs which represent the
301  * path of indirect blocks required to access a data block.  The first "pair"
302  * contains the logical block number of the appropriate single, double or
303  * triple indirect block and the offset into the inode indirect block array.
304  * Note, the logical block number of the inode single/double/triple indirect
305  * block appears twice in the array, once with the offset into the i_ib and
306  * once with the offset into the page itself.
307  */
308 int
309 ext2_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump)
310 {
311 	long blockcnt;
312 	e2fs_lbn_t metalbn, realbn;
313 	struct ext2mount *ump;
314 	int i, numlevels, off;
315 	int64_t qblockcnt;
316 
317 	ump = VFSTOEXT2(vp->v_mount);
318 	if (nump)
319 		*nump = 0;
320 	numlevels = 0;
321 	realbn = bn;
322 	if ((long)bn < 0)
323 		bn = -(long)bn;
324 
325 	/* The first EXT2_NDADDR blocks are direct blocks. */
326 	if (bn < EXT2_NDADDR)
327 		return (0);
328 
329 	/*
330 	 * Determine the number of levels of indirection.  After this loop
331 	 * is done, blockcnt indicates the number of data blocks possible
332 	 * at the previous level of indirection, and EXT2_NIADDR - i is the
333 	 * number of levels of indirection needed to locate the requested block.
334 	 */
335 	for (blockcnt = 1, i = EXT2_NIADDR, bn -= EXT2_NDADDR; ;
336 	    i--, bn -= blockcnt) {
337 		if (i == 0)
338 			return (EFBIG);
339 		/*
340 		 * Use int64_t's here to avoid overflow for triple indirect
341 		 * blocks when longs have 32 bits and the block size is more
342 		 * than 4K.
343 		 */
344 		qblockcnt = (int64_t)blockcnt * MNINDIR(ump);
345 		if (bn < qblockcnt)
346 			break;
347 		blockcnt = qblockcnt;
348 	}
349 
350 	/* Calculate the address of the first meta-block. */
351 	if (realbn >= 0)
352 		metalbn = -(realbn - bn + EXT2_NIADDR - i);
353 	else
354 		metalbn = -(-realbn - bn + EXT2_NIADDR - i);
355 
356 	/*
357 	 * At each iteration, off is the offset into the bap array which is
358 	 * an array of disk addresses at the current level of indirection.
359 	 * The logical block number and the offset in that block are stored
360 	 * into the argument array.
361 	 */
362 	ap->in_lbn = metalbn;
363 	ap->in_off = off = EXT2_NIADDR - i;
364 	ap++;
365 	for (++numlevels; i <= EXT2_NIADDR; i++) {
366 		/* If searching for a meta-data block, quit when found. */
367 		if (metalbn == realbn)
368 			break;
369 
370 		off = (bn / blockcnt) % MNINDIR(ump);
371 
372 		++numlevels;
373 		ap->in_lbn = metalbn;
374 		ap->in_off = off;
375 		++ap;
376 
377 		metalbn -= -1 + off * blockcnt;
378 		blockcnt /= MNINDIR(ump);
379 	}
380 	if (nump)
381 		*nump = numlevels;
382 	return (0);
383 }
384